Remove AI from this picture and it says, “Participating in the public sphere is risky. Someone could hurt you or trick you. They could listen to what you say and use it against you. If you openly admit what you value, someone could hold it for ransom to compel your cooperation in their schemes.”
Which is all true.
But it’s also true that participating in the public sphere enables cooperation; enables mutual aid; enables creating reputation and credibility; enables joining with others to achieve your values.
But it’s also true that participating in the public sphere enables cooperation; enables mutual aid; enables creating reputation and credibility; enables joining with others to achieve your values.
Before internet, we had “public spheres” of various size—there is a difference between e.g. telling something to a group of friends, speaking at a village meeting, or publishing a book.
Internet kinda makes it all one size—anything said anywhere could leak anywhere else. Which changes the equation: you can’t choose smaller sphere for smaller risks and smaller benefits; now everything is associated with the large risks. You can’t trust that things said to your friends won’t reach your employer.
Internet kinda makes it all one size—anything said anywhere could leak anywhere else. Which changes the equation: you can’t choose smaller sphere for smaller risks and smaller benefits; now everything is associated with the large risks. You can’t trust that things said to your friends won’t reach your employer.
This seems wrong to me? Discord, text messages, slack, invite-only forums, private social media accounts, and niche forums are all different levels of community size, and while for all of those there is some risk of what you say spreading to everyone you know, and that risk is larger than if you only communicated in person to a single individual at a time, I’m not convinced that that risk is uniform or significant across all mediums. One need only use common sense.
A village meeting for instance is likely more risky on this front too than a post on a niche forum, despite the latter technically being accessible to the whole world.
Agreed, the risk is not entirely uniform. The risk of a screenshot leaking is real—without thinking I immediately remembered two cases from the rationalist community; in one case, someone exposed opinions posted in a private group chat; in another case someone exposed a privately sent e-mail; in both cases the intent was to expel someone from the community because of their political opinions; in both cases the author trusted the recipient(s) to be friend(s) and said things they normally would not publish (but nothing horrible, in my opinion).
Compared to the offline world, I think if someone said a similar thing in a pub, and someone quoted them, they could simply deny it. And if someone recorded and published a private conversation in a pub, that would be an obvious breach of etiquette; I don’t think that person would ever again be invited to the pub or to any private meeting, not even by members of the same political faction.
Alice sends Bob an email containing text T, and Bob tells Charlie “Alice said T″!” and sends a screenshot of the email. Charlie, convinced, sanctions Alice for saying T.
Alice tells Bob a statement T in a pub, and Bob tells Charlie “Alice said T″!”. However, Alice hears that Bob told Charlie this, and denies she ever said T. Charlie believes Alice, or at least is left with enough uncertainty not to sanction Alice for saying T.
I guess I don’t understand why Alice can’t deny having sent Bob T in an email in situation 1. Bob could have made up the text just as much as he could have made up the statement in 2, especially now, in the age of LLMs. Maybe Alice also sent a signature cryptographically verifying they sent T, however simply not sending that signature seems easier than abandoning email as a medium for this type of communication.
Edit: There are also other solutions here, like automatically deleting messages before webscrapers can touch them, or using the anonymity still by default provided by the internet.
AIs will get more powerful & very significant pretty soon, it seems pretty relevant to ask whether one should take actions to prevent AIs from knowing about you, and pretty plausible to suggest that while the particular problems with being known publicly will stick around, they may get worse, and we may get a very different distribution over upsides and downsides.
I think your comment ignores the fact that OP is stating what problems (and benefits) they expect to be more salient if an AI is able to train on your posts.
Remove AI from this picture and it says, “Participating in the public sphere is risky. Someone could hurt you or trick you. They could listen to what you say and use it against you. If you openly admit what you value, someone could hold it for ransom to compel your cooperation in their schemes.”
Which is all true.
But it’s also true that participating in the public sphere enables cooperation; enables mutual aid; enables creating reputation and credibility; enables joining with others to achieve your values.
Before internet, we had “public spheres” of various size—there is a difference between e.g. telling something to a group of friends, speaking at a village meeting, or publishing a book.
Internet kinda makes it all one size—anything said anywhere could leak anywhere else. Which changes the equation: you can’t choose smaller sphere for smaller risks and smaller benefits; now everything is associated with the large risks. You can’t trust that things said to your friends won’t reach your employer.
This seems wrong to me? Discord, text messages, slack, invite-only forums, private social media accounts, and niche forums are all different levels of community size, and while for all of those there is some risk of what you say spreading to everyone you know, and that risk is larger than if you only communicated in person to a single individual at a time, I’m not convinced that that risk is uniform or significant across all mediums. One need only use common sense.
A village meeting for instance is likely more risky on this front too than a post on a niche forum, despite the latter technically being accessible to the whole world.
Agreed, the risk is not entirely uniform. The risk of a screenshot leaking is real—without thinking I immediately remembered two cases from the rationalist community; in one case, someone exposed opinions posted in a private group chat; in another case someone exposed a privately sent e-mail; in both cases the intent was to expel someone from the community because of their political opinions; in both cases the author trusted the recipient(s) to be friend(s) and said things they normally would not publish (but nothing horrible, in my opinion).
Compared to the offline world, I think if someone said a similar thing in a pub, and someone quoted them, they could simply deny it. And if someone recorded and published a private conversation in a pub, that would be an obvious breach of etiquette; I don’t think that person would ever again be invited to the pub or to any private meeting, not even by members of the same political faction.
Hm. So the two situations are:
Alice sends Bob an email containing text T, and Bob tells Charlie “Alice said T″!” and sends a screenshot of the email. Charlie, convinced, sanctions Alice for saying T.
Alice tells Bob a statement T in a pub, and Bob tells Charlie “Alice said T″!”. However, Alice hears that Bob told Charlie this, and denies she ever said T. Charlie believes Alice, or at least is left with enough uncertainty not to sanction Alice for saying T.
I guess I don’t understand why Alice can’t deny having sent Bob T in an email in situation 1. Bob could have made up the text just as much as he could have made up the statement in 2, especially now, in the age of LLMs. Maybe Alice also sent a signature cryptographically verifying they sent T, however simply not sending that signature seems easier than abandoning email as a medium for this type of communication.
Edit: There are also other solutions here, like automatically deleting messages before webscrapers can touch them, or using the anonymity still by default provided by the internet.
AIs will get more powerful & very significant pretty soon, it seems pretty relevant to ask whether one should take actions to prevent AIs from knowing about you, and pretty plausible to suggest that while the particular problems with being known publicly will stick around, they may get worse, and we may get a very different distribution over upsides and downsides.
I think your comment ignores the fact that OP is stating what problems (and benefits) they expect to be more salient if an AI is able to train on your posts.